Machine learning identifies routine blood tests as accurate predictive measures of pollution-dependent poor cognitive function
4 affiliations
- 1School of Geosciences, University of Aberdeen, Aberdeen, UK
- 2Theoretical Biophysics, Institute for Biology, Humboldt University of Berlin, Berlin, Germany
- 3Institute of Medical Sciences, School of Medicine, Medical Sciences and Nutrition, University of Aberdeen, Foresterhill, Aberdeen, AB25 2ZD UK
- 4Aberdeen Biomedical Imaging Centre, School of Medicine, Medical Sciences and Nutrition, University of Aberdeen, Aberdeen, UK
Abstract
Several modifiable risk factors for dementias have been identified including socio-economic status and environmental exposures – however, how these population-level risks relate to individual risk remains elusive. To address this, we use random forest modelling to determine significant predictors of poor cognitive performance in on deeply phenotyped cohort of 324 individuals (age 61.6 ± 4.8 years; 150 males, 174 females) without extant neurological disease. 457 features were assessed including a comprehensive battery of imaging, blood, atmospheric pollutant and socio-economic measures. This approach where brain imaging, blood, socio-economic and environmental exposure measures were used to model cognitive performance was able to classify general cognition tertiles with 0.70 overall accuracy. Routinely assessed markers of anaemia including mean corpuscular haemoglobin were identified as predictors of poor general cognition and both extremes (low and high) of mean corpuscular haemoglobin concentration were associated with poor immediate recall (p = 0.0253). The predictors of poor cognition which consistently improved model performance across all models were measures of atmospheric pollution, in particular, lead, carbon monoxide, and particulate matter. Feature analysis demonstrated a significant negative relationship between low mean corpuscular haemoglobin concentration and high levels of atmospheric pollutants (p < 0.05) highlighting the potential of routinely assessed blood tests as predictive measures of pollution-dependent cognitive functioning, at an individual level. These data demonstrate how routine medical testing and local authority initiatives could identify at-risk individuals, highlighting the potential for targeted, cost-effective medical and social interventions to improve population cognitive health.
Supplementary Information: The online version contains supplementary material available at https://doi.org/10.1038/s41598-026-64389-9.
Related
- Stage-Stratified Explainable Machine Learning Identifies a Reproducible Transcriptomic Classifier with Stage-Dependent Feature Importance in Colorectal Cancer: A Multi-Cohort Study with Saudi Population Contextualisation
- Explainable machine learning identifies predictors for postural instability and cognitive impairment in early Parkinsons Disease
- Association between PCSK9 and post-stroke cognitive impairment: a retrospective cohort study and predictive models establishment based on machine learning.
- Air pollution is associated with sex‐specific longitudinal changes in astrocytic and neuroinflammatory processes linked to cognitive decline in Alzheimer's disease at‐risk individuals
- Serum Proteomic Profiling by SomaScan Aptamer Array Identifies Candidate Predictive and Monitoring Biomarkers of Distant Recurrence in Triple-Negative Breast Cancer
BibTeX
@article{johnson2026machine,
title = {{Machine learning identifies routine blood tests as accurate predictive measures of pollution-dependent poor cognitive function}},
author = {Hamish Johnson and James Longden and Gary Cameron and Gordon D. Waiter and Fergal M. Waldron and Jenna M. Gregory and Holly Spence},
journal = {Scientific Reports},
year = {2026},
month = jul,
doi = {10.1038/s41598-026-64389-9},
url = {https://doi.org/10.1038/s41598-026-64389-9},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Machine learning identifies routine blood tests as accurate predictive measures of pollution-dependent poor cognitive function
AU - Hamish Johnson
AU - James Longden
AU - Gary Cameron
AU - Gordon D. Waiter
AU - Fergal M. Waldron
AU - Jenna M. Gregory
AU - Holly Spence
PY - 2026
DA - 2026/07/29
JO - Scientific Reports
DO - 10.1038/s41598-026-64389-9
UR - https://doi.org/10.1038/s41598-026-64389-9
AB - Several modifiable risk factors for dementias have been identified including socio-economic status and environmental exposures – however, how these population-level risks relate to individual risk remains elusive. To address this, we use random forest modelling to determine significant predictors of poor cognitive performance in on deeply phenotyped cohort of 324 individuals (age 61.6 ± 4.8 years; 150 males, 174 females) without extant neurological disease. 457 features were assessed including a comprehensive battery of imaging, blood, atmospheric pollutant and socio-economic measures. This approach where brain imaging, blood, socio-economic and environmental exposure measures were used to model cognitive performance was able to classify general cognition tertiles with 0.70 overall accuracy. Routinely assessed markers of anaemia including mean corpuscular haemoglobin were identified as predictors of poor general cognition and both extremes (low and high) of mean corpuscular haemoglobin concentration were associated with poor immediate recall (p = 0.0253). The predictors of poor cognition which consistently improved model performance across all models were measures of atmospheric pollution, in particular, lead, carbon monoxide, and particulate matter. Feature analysis demonstrated a significant negative relationship between low mean corpuscular haemoglobin concentration and high levels of atmospheric pollutants (p < 0.05) highlighting the potential of routinely assessed blood tests as predictive measures of pollution-dependent cognitive functioning, at an individual level. These data demonstrate how routine medical testing and local authority initiatives could identify at-risk individuals, highlighting the potential for targeted, cost-effective medical and social interventions to improve population cognitive health. Supplementary Information: The online version contains supplementary material available at https://doi.org/10.1038/s41598-026-64389-9.
SN - 2045-2322
ER -